Search used to end with a page of options and a human deciding. Increasingly it ends with a synthesised answer and a human accepting it.
That single change breaks an assumption the entire SEO industry was built on: that being visible in the results is the same as being considered.
Two jobs, not one
Search now has two jobs, and most companies are only doing the first.
The first is the familiar one: rank for the queries that matter. Still worth doing, still where a lot of qualified traffic comes from.
The second is newer. When someone asks an assistant which vendors to look at in your category, you want to be one of the names that comes back. That is a different mechanic. It is not about ranking a page, it is about whether a model has enough consistent, corroborated information to name you confidently.
What makes a model name you
From what I see working, three things move the needle.
Being described consistently everywhere. If your own site, your directory listings, your LinkedIn and any press all describe what you do in materially different words, you are hard to summarise. Models resolve ambiguity by picking whoever is unambiguous.
Being mentioned somewhere other than your own site. Self-description is weak evidence. Third-party mentions are what make a claim about you look corroborated.
Answering questions in the shape they are asked. Pages built around a real question, with the answer near the top and the reasoning below, are far easier to quote than a page built around a keyword.
That last one is why FAQ sections stopped being filler and became infrastructure.
The uncomfortable part for content teams
This rewards a kind of writing most marketing departments are structurally bad at: committing to an answer.
Hedged content, the sort that lists every consideration and concludes that it depends, is unquotable. There is nothing to extract. Content with a clear position, supported by a specific reason, is extremely quotable, which is exactly why it also performs better with humans.
The optimisation and the quality argument point the same way for once.
What I would do first
If you have limited time, I would not start with technical work. I would start by writing five pages that answer the five questions your buyers actually ask before they shortlist, with your genuine opinion in each, and make sure your description of yourself is identical across every property you control.
Technical structure and schema help, but they help a page that has something to say. They cannot help a page that does not.
FAQ
Is traditional SEO dead?
No. Ranking still drives meaningful traffic and it feeds the sources assistants read. What changed is that ranking is no longer sufficient, because a growing share of queries never produce a click.
How do I know if AI assistants mention my brand?
Ask them, repeatedly and in the way a buyer would phrase it, and record the answers over time. It is manual, but it is the only measurement that reflects the actual buying moment.
Does adding an FAQ section really help?
Yes, for a practical reason: a question with a direct answer underneath is the easiest possible unit to extract and quote. It also forces you to state a position rather than describe a topic.
Should I block AI crawlers to protect my content?
Only if being cited has no value to you, which is rare in B2B. If you sell to people who ask assistants for recommendations, being absent from the training and retrieval layer is the expensive option.
